• Do not lock a table that is involved in import processing. If such a table is locked, import processing might result in the lock release wait status or deadlock.
• If a table at a target system is updated from multiple source systems, the order of the update processing cannot be predicted. If you need to control the order of update processing, you must do so by handling the processing appropriately. Suppose that data linkage is established from two source systems, A and B, to table T1 at a target system, and that a table row corresponding to source system A's mapping key value 100 is updated. Then a table row corresponding to source system B's mapping key value 100 is updated. In this case, the contents of the table row corresponding to key value 100 at the target system might not be the same as that of the corresponding row at source system B.
• Before you modify the extraction definition at the source system, you must ensure that all update information has been sent from the source system to the target Datareplicators. If you modify the extraction definition while some update information still remains untransmitted at the source Datareplicator, import processing might result in an error. In such a case, a target Datareplicator might not be able to recover the error, necessitating use of a program such as HiRDB Datareplicator to re-create tables.
• The table-based import method is always used when a UOC routine is used for import processing.

• To initialize the import environment for a specified data linkage identifier only, you must execute partial initial start. To initialize the import environments for all data linkage identifiers, terminate import processing for all data linkage identifiers, and then execute initial start on the entire import environment.
• If you start the source system in the initialized status, start the target Datareplicator with the hdsstart -i or hdsstart -i -D command. If you start only the source system in the initialized status, the KFRB02003-E error results (detail code 5).
  If you specified nocheck in the extract_init operand in the import environment definition, there is no need to start the target Datareplicator in the initial start or partial initial start mode.
  

• Even if an event code sent from the source system is undefined in the import environment definition, there is no effect on the import processing; the target Datareplicator still recognizes it as an event.
• While import processing is underway, the target Datareplicator ignores any transaction-based or table-based import restart event that is detected.

The following prerequisites and limitations apply to data linkage on tables for which the WITHOUT ROLLBACK option is specified.

Prerequisites:

The prerequisites listed below must all be satisfied. If any of the prerequisites is not satisfied, an inconsistency might occur between the source and target databases.

• The versions of the source and target Datareplicators are both 08-02 or later.
• The source and target databases are both HiRDB.
• nodemst is specified in the sendcontrol operand in the extraction system definition.
• If the source and target databases are on a parallel server, the table for which the WITHOUT ROLLBACK option is specified and all tables related to that table are stored on the same back-end server.

Limitations:

The following limitations apply to data linkage on tables for which the WITHOUT ROLLBACK option is specified:

• Data linkage is not supported for the following tables when the WITHOUT ROLLBACK option is specified:
  Merge tables
  Time-ordered information tables
  Tables whose rows are partitioned among servers
  Event control tables
  
• Only the UPDATE SQL statement can be used to perform data linkage. Because the INSERT statement cannot be used for data linkage, use a program such as HiRDB Dataextractor to create the target database.
• If the WITHOUT ROLLBACK option is specified for a table, its column specified for mapping keys cannot be updated.
• Import errors cannot be skipped. Specify SQL coding to skip such errors.
• The data linkage recovery facility is not supported. Recover data linkage errors individually.

(a) Notes about referencing the target database

For a table for which the WITHOUT ROLLBACK option is specified, one update operation is treated as one transaction. Therefore, if a table for which the WITHOUT ROLLBACK option is specified and a table for which the option is omitted are updated within the same transaction, the target Datareplicator sends and imports data as separate transactions. If the target database is referenced before all the transactions have been imported, an inconsistency might occur between the source and target databases. Therefore, make sure that the import processing has been completed^# before you reference the target database.

Furthermore, if the target Datareplicator rolls back processing due to a failure, the contents of a table for which the WITHOUT ROLLBACK option is specified might not be refreshed temporarily while update information is being imported. Therefore, make sure that the import processing has been completed^# before you reference the target database.

#

Make sure that read position and write position values that are displayed by executing the hdsstate are the same.

(b) Specification of the sendintvl operand in the transmission environment definition

For tables for which the WITHOUT ROLLBACK option is specified, one update operation is treated as one transaction. If 0 (send update information in units of transactions) is specified in the sendintvl operand in the transmission environment definition, transmission processing occurs each time a table for which the WITHOUT ROLLBACK option is specified is updated, resulting in a large overhead for transmission processing. Therefore, if you apply data linkage to tables for which the WITHOUT ROLLBACK option is specified, we recommend that you specify a nonzero value in the sendintvl operand in the transmission environment definition.

The following prerequisites apply to data linkage on tables for which the COMPRESSED option is specified.

Prerequisites:

The prerequisites listed below must all be satisfied. If any of the prerequisites is not satisfied, an inconsistency might occur between the source and target databases or import processing might result in an SQL execution error.

• The versions of the source and target Datareplicators are both 08-06 or later.
• The source and target databases are both HiRDB.
• The same endian is used for both the source and the target database.
• The compression specification is the same between the source and target tables.